Nvidia Has Risk — But It’s Not What You Think
The US should lift all restrictions on Nvidia selling into China
Michael Burry has built himself a very comfortable business. With thousands of paid Substack subscribers, he’s making a handsome paycheck doing what he does best: holding up a scary mirror to anxious investors. His latest target is Nvidia, and his critique tells you two things:
He’s a professional scaremonger.
He’s touting the wrong risk.
Burry has no skin in the game, just a subscription business that thrives on tension. His model is simple: take a wildly successful company, sprinkle some accounting jargon, invoke the ghost of the dot-com bust, and watch the checks roll in. It works because Nvidia has been the best-performing stock in the S&P 500 for years, leaving a lot of investors sitting on massive unrealized gains and quietly terrified of the inevitable crash everyone keeps predicting.
And Burry’s core argument? That hyperscalers are depreciating GPUs over five or six years while the chips become obsolete in two or three, setting the stage for gigantic future write-downs. Translation: “If demand collapses, earnings collapse.”
Well, duh.
That’s not analysis; it’s a tautology. It’s like standing at the bottom of a 50-foot wave and shouting, “If the surfer falls, he’s going to wipe out hard!” Yes, Michael, we know. The interesting question is whether the rider makes the wave or not.
The real risk to Nvidia is not that demand suddenly evaporates. The real risk is the opposite: that demand keeps exploding and misguided U.S. policy hands the entire market to China.
Real-world AI — autonomous vehicles, robotics, scientific simulation, next-gen video-game AI like Elon’s promised Grok-5 that can watch pixels and play League of Legends at superhuman level — requires training on 4D data (3D + time). That is orders of magnitude more compute- and memory-intensive than text. Even the so called multimodal models available today (Gemini, Grok, GPT etc.) compress everything to a one dimensional data string. Perceiving and analyzing the world in 4D is a whole new level of compute. Nvidia is the only company that is obsessively building the full-stack hardware (compute, memory, networking, software) optimized for exactly that workload. CEO Jensen Huang is literally betting the company on it.
Meanwhile, China is pouring tens of billions into domestic alternatives because no government wants its entire AI future subject to the whims of Washington export controls.
If the U.S. keeps tightening restrictions, Nvidia’s position in the global market will be at risk. Ironically, this could accelerate the outcome Burry claims to have discovered: that Nvidia actually becomes Cisco — a once-dominant franchise that looses to Asian competition.
So the biggest risk to Nvidia isn’t overbuild or depreciation schedules.
It’s that we forget how to compete and instead choose to hide.
Pundits like Burry are entertainers, not investors. Their product is fear, and fear sells subscriptions. The adults in the room should focus on the actual threat: making sure American companies — starting with Nvidia — can sell freely into the largest and most motivated compute market on the planet.
That’s how you keep riding the wave instead of watching China surf it without us.